University of Groningen Vietnamese learners mastering english articles Thu, Huong Nguyen; Huong, N.T. IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record Publication date: 2005 Link to publication in University of Groningen/UMCG research database Citation for published version (APA): Thu, H. N., & Huong, N. T. (2005). Vietnamese learners mastering english articles s.n. Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons). Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum. Download date: 02-12-2017
Chapter 2 ENGLISH ARTICLES: A Problem for Vietnamese Students? The purpose of the present chapter is to see whether Vietnamese students have problems with English articles and if so, how these errors relate in number to other kinds of errors, whether the number of article errors decreases as proficiency increases, and finally whether these errors affect their writing grades. So far many researchers have agreed that article usage is an area of grammar that shows considerable prominence in any error analysis (Oller & Redding, 1971: 85). Especially studies focusing on Asian learners of English as a foreign language (EFL), namely Japanese and Chinese, have reported a high frequency of article errors among common errors in English (for an overview see Goto Butler, 2002). However, as far as we know, no study has shown that article errors are indeed relatively more frequent for Asian students than other types of errors. In our study, we will first look at the total number of grammar errors and then see how relatively frequent article errors are. Oller & Redding (1971), who used a multiple-choice format to test mastery of the English article system, found that article errors decrease as proficiency increases, even for students whose L1 does not have formal equivalents of the English articles. In our study, we will perform correlation tests on students proficiency scores and numbers of various errors that students make to find out if the number of article errors indeed decreases as students become more proficient. Chapter 2 7
The studies reviewed in Goto Butler (2002) are based on different ways of collecting data, ranging from personal experience in teaching to observational and empirical evidence. Tasks used for collecting data also vary from recognition tasks (e.g. multiple-choice, error identification) to production tasks (e.g. composition, gap-filling, interviews). Researchers have commented on a possible relationship between the frequency of each error type and the task or test performed (cf. Ellis 1994, pp138). In general, discrete point tests such as cloze tests provide higher error rates than production tests like essay writing and interviews. The reason for the difference, according to researchers, lies in the fact that students tend to avoid certain uses they are not sure of in production tasks. In our study, essay writing, a type of production task, is used instead of a discrete point test because we want to see if article errors affect writing grades. Also, because almost every noun phrase produced involves choices in article use, the assumption is that article use cannot be avoided and article errors will be relatively frequent, even in written production. However, the correct or incorrect use of articles does not usually affect understanding, and as Burt (1975) points out does not necessarily affect communication. The assumption, therefore, is that if article errors affect grades, they will do so only to a small degree. DATA COLLECTION To determine the relationship between article errors, essay grades and proficiency scores in high-intermediate Vietnamese students of English, we tested 14 junior and senior university students of English as a foreign language, trained to become teachers of English in senior high schools, at Cantho University in Vietnam. Junior and senior students (i.e. 3 rd year and Chapter 2 8
4 th year students) were selected instead of freshman and sophomore students (i.e. 1 st and 2 nd year) because they were assumed to be at a high-intermediate level and had already taken courses in Practical English Grammar. The data collection was carried out during the regular teaching term. Via announcements in classes, students were asked to volunteer for this experiment. Twenty-nine students originally participated, but only 14 of those were selected based on their proficiency scores. Subjects who had obtained a score of less than 38 out of 63 were eliminated from the study, for it was assumed that the lower scores were no better than chance and therefore unreliable. Students were asked to write an English essay of about 400 words within 60 minutes. The essay topics, consisting of two options, were drawn from a commercial TOEFL test preparation book. (See Appendix 1.) To test proficiency, the 2001 Michigan Proficiency test was used. (See Appendix 3.) DATA ANALYSIS AND RESULTS Question 1: How frequent are article errors? To determine the frequency of occurrence of article errors and the relationship between article errors and other types of errors, two teachers (the author and an instructor of English at the university of Groningen) graded the 14 essays on grammar errors exclusively. After the two raters had each marked each essay separately and categorized each error, they compared errors and categories and resolved differences in consultation. The errors were categorized globally into syntactic, semantic and mechanical errors. The syntactic errors were further sub-categorized Chapter 2 9
according to the place of occurrence: in the noun phrase (NP), the verb phrase (VP), or clause structure (CS). Lexical errors (LE) were considered semantic errors, and errors in punctuation, capitalization and so on were considered mechanical (ME) errors. See Table 1 for more detail on each category. Each error was counted separately, even if a similar error was made more than once. (See Appendix 2.) To be able to compare across essays, the errors were calculated relative to the number of words in the essays. Spearman s rho nonparametric correlations were performed to find out the relationship between the different categories of errors. Table 1. Types and relative frequency of form errors in essays Error types Descriptions of errors Number of errors Percentage n=378 Noun phrase errors Incorrect use of articles; Faulty premodification 135 35.7% or post-modification Verb phrase errors Incorrect use of tenses; verb forms; voice; mood; aspect; agreement. 38 10.0% Clause structure Lack of subject; subcategorization error; 62 16.4% errors incorrectly formed constituent; wordorder problem Lexical errors Word choice and word form 72 19.0% Mechanics Spelling; Punctuation; Lack of 71 18.8% consistency in pronoun use; Fragments; Comma splice; Fused sentences. In total 378 errors were made; with 35.7%, the NP errors accounted for the highest percentage of errors. Within this category, the most frequent error concerned English articles (88.2%). Therefore article errors accounted for 31.5% of all errors made in the essays. Since our interest is in the frequency of article errors in comparison with other types of errors, we will first analyze the relation between the total number of errors with article errors and other major types of errors and then Chapter 2 10
that between the article errors and other major types of errors. First, as far as the relation between the total number of errors and other types of errors is concerned, the results from the correlation test in Table 2 show a significantly positive relation between the total number of errors and such major error types as verb phrase, clause, and mechanical ones. No significantly positive relation was found between the total number of errors and those errors relating to noun phrase, word use (i.e. lexis), and namely article. It is worth noting that the positive relation between these error types and the total number of errors was not high. Table 2. Correlations among the errors (n=14) (2-tailed) Relative number of errors Article Noun Phrase Verb phrase Clause Word use Mechanics Relative number of errors Corr. coeffi. Sig. 1.000..472.088.486.078.767.001.569.034.433.122.640.014 Article Corr..472 1.000.911.299 -.180 -.357.193 coeffi. Sig..088..000.298.539.210.509 Noun Phrase Corr. coeffi..486.911 1.000.336 -.262 -.363.196 Sig..078.000..240.366.203.503 Verb Phrase Corr. coeffi..767.299.336 1.000.288.218.793 Sig..001.298.240..318.455.001 Clause Corr..569 -.180 -.262.288 1.000.538.363 coeffi. Sig..034.539.366.318..047.203 Word use Corr..433 -.357 -.363.218.538 1.000.165 coeffi. Sig..122.210.203.455.047..573 Mechanic Corr..640.193.196.793.363.165 1.000 coeffi. Sig..014.509.503.001.203.573. Second, regarding the correlation between article errors and other types of errors, no significant correlation was found between article errors and those of others except that between the article and noun phrase errors, Chapter 2 11
which was significantly positive. This indicates that if there is a high level of errors in using noun phrases, there will be a high level of errors in using articles. Question 2: Do the relative frequency of article errors affect writing grades? To determine whether or not article use errors influence essay grades, each of the 14 essays was graded on a scale of 1-10. Five university teachers of English at Cantho University (all native Vietnamese speakers) and two teachers at the University of Groningen (one native Dutch speaker and one native English speaker) were asked to grade each essay holistically on the basis of content and form. Table 3 (p. 14) shows that there was no significant relation between teachers scores (i.e. WM) and the relative number of article errors (r=-.226). This revealed that there were essays receiving high scores although they had many article errors (see Fig.1). In addition, the table also presents a somewhat different picture when a distinction is made between the scores of two groups of teachers (i.e. Dutch and Vietnamese) in relation to the article errors. Dutch teachers seemed to have a tendency (r= -.455) to take into account the article errors (although the correlation was not significant). Vietnamese teachers, by contrast, did not pay any attention to the problem (r= -.029) through the fact that there was no correlation between their scores and the article errors. It is also interesting to note that essay scores correlated significantly and negatively with the relative number of errors found in clauses, verb phrases, and word use. No significant relations were found between grading and either mechanical errors or noun phrase ones. A further analysis into each group of teachers in Table 3 also revealed that teachers in two groups Chapter 2 12
focused much on the errors regarding verb phrases, which contains the core message of a sentence, rather than on noun phrase ones. 30 20 10 DETERMI 0 3 4 5 6 7 8 WMEAN Fig. 1. Scattergram showing relationships between grading and article errors. Chapter 2 13
VT DT WM ART Table 3. Correlation coefficients of holistic writing grades and article use errors (N=14; 2-tailed) VT DT WM RelErr. CS VP ME LE NP ART Corr. Coeffi 1.000.568.938 -.543 -.446 -.618 -.310 -.618.055 -.029 Sig...034.000.045.110.019.281.019.852.923 Corr. Coeffi.568 1.000.762 -.822 -.517 -.575 -.508 -.422 -.462 -.455 Sig..034..002.000.058.032.063.133.096.102 Corr. Coeffi.938.762 1.000 -.719 -.534 -.692 -.371 -.587 -.138 -.226 Sig..000.002..004.049.006.191.027.637.436 Corr. Coeffi.029 -.455 -.226.472 -.191.407.235 -.398.956 1.000 Sig..923.102.436.088.513.149.418.159.000. VT is average of grades given by Vietnamese teacher DT is average of grades given by Dutch teachers score WM is average of grades given by Vietnamese and Dutch teachers Rel.Err. is the relative number of the total number of errors in the essays CS is the relative number of clause errors VP is the relative number of verb phrase errors ME is the relative number of mechanical errors LE is the relative number of word use errors NP is the relative number of noun phrase errors ART is the relative number of article errors
Question 3: Is there a correlation (a) between essay grades and proficiency scores and (b) between proficiency scores and article errors and errors of other types? To test proficiency, the 2001 Michigan Proficiency Test was administered to the students. The test consisted of 50 items focusing on grammar, 27 items on vocabulary meaning, and 23 items on vocabulary in context. All the items were designed in the form of multiple-choice questions. Students took the proficiency test in 60 minutes. The test was graded on the basis of a correct/incorrect scoring method. Spearman s non-parametric correlations were conducted to find out the relationship between proficiency scores and essay grades and between the number of article errors and proficiency level. First, concerning proficiency scores and essay grades, no significant correlation was found although the relation had a positive tendency, suggesting that students have high essay scores if they are proficient. Given that there was a tendency, the mean writing grades of Vietnamese and Dutch teachers did not significantly correlate with proficiency scores (r=.466; p=.093). However, when grades by the Dutch and Vietnamese teachers are taken separately, we find a positive relationship between Dutch teachers essay grades and proficiency (r=.604; p=.022), but not between Vietnamese teachers essay grades and proficiency scores (Table 4). Second, as far as the relation between proficiency scores and article errors is concerned, no significant correlation was found (r=-.433; p=.122; 2-tailed). Yet, the relation had a negative tendency. This means that about 16 percent of the variation among article errors is explained by proficiency scores. This goes against what Oller et al. (1971) confirmed that there was a close relation between skill in article usage and overall proficiency. Chapter 2 15
Table 4. Spearman correlations between article use errors and grading and proficiency scores (n=14; 2-tailed) Profi. CS ME VP LE NP ART WM VT DT Profi. Corr. Coef. 1.000 -.188 -.053 -.241.201-396 -.433.466.294.604 Sig...520.857.407.491.161.122.093.308.022 ART Corr. Coef. -.433 -.191.235.407 -.398.956 1.000 -.226 -.029 -.455 Sig..122.513.418.149.159.000..436.923.102 WM Corr. Coef.466 -.534 -.371 -.692 -.587 -.138 -.226 1.00.938.762 Sig..093.049.191.006.027.637.436..000.002 VT Corr..294 -.446 -.310 -.618 -.618.055 -.029.938 1.00.568 Coef Sig..308.110.281.019.019.852.923.000..034 DT Corr..604 -.517 -.508 -.575 -.422 -.462 -.455.762.568 1.00 Coef Sig..022.058.063.032.133.096.102.002.034. Profi is proficiency scores CS is the relative number of clause errors ME is the relative number of mechanical errors VP is the relative number of verb phrase errors LE is the relative number of word use errors NP is the relative number of noun phrase errors ART is the relative number of article errors WM is average of grades given by Vietnamese and Dutch teachers VT is average of grades given by Vietnamese teacher DT is average of grades given by Dutch teachers score
Third, concerning the relation between other error types and proficiency scores, no significant correlation was found. From these correlation analyses, although no significant correlations were found, it is worth noting that careful interpretations should be taken into account given the fact that the sample is not large enough. Question 4. Do essay grades improve when article errors are eliminated? To determine whether or not grades improve when article errors are removed, four essays out of the 14 graded were selected (Tables 5 & 6), representative of the essays with the most article errors (13 or more) and the least article errors (4 or less). Two essays had high grades (over 6.8 average on a scale of 1-10) and two had low grades (less than 4.2 average). Table 5. Selection of essays for question 3 Essay Writing grade (1-10) Number of article errors A 4.2 4 B 6.9 3 C 6.8 23 D 3.2 13 From this set of essays another set was created in which the article errors were corrected. Thus, in total there were eight essays, four original ones and four corrected ones. Twenty advanced University Dutch students of English graded these essays. Each student received four different essays, two of which had the article errors corrected and two uncorrected. The corrected and uncorrected essays were systematically varied. Chapter 2 17
Table 6. Means scores of the four essays selected for question 3 Essays Number of Std Mean N Article errors Deviation A 4 Original 4.86 11.8970 0 Corrected 5.11 9.8937 B 3 Original 6.45 10.4378 0 Corrected 6.65 10 1.1797 C 23 Original 6.5 10.8498 0 Original 6.8 10 1.0328 D 13 Original 5.2 9.8700 0 Corrected 5.6 11.9700 With respect to the relationship between grading and the removal of article errors, the results obtained from the ANOVA computation indicate that there are significant differences among students and among the four essays (see Table 7), but not between corrected and uncorrected essays. Table 7. Analysis of variance (ANOVA) Df F Sig. Students 19 1.810.045 Essays 3 20.512.000 Correction 1 2.319.133 DISCUSSION This study shows that article errors occur relatively frequently (31.49%) when compared to other errors such as errors in the verb phrase, the clause, lexis, mechanics, or sentence. Therefore, it is evident that correct English article use is a problem for Vietnamese students, even for high-intermediate university students who are about to become teachers of English. In addition, the fact that there is no significant relationship between the number of article errors and the total number of errors found in the essays (r=.472; p=.088) supports the idea that even good essays (i.e. essays with Chapter 2 18
a low total number of errors) may still have a relatively high degree of article errors. This study also shows that there is no correlation between essay grades and the number of article errors (Table 4). However, when the mean grades of Vietnamese teachers and Dutch teachers are considered separately, a slightly different picture emerges. Whereas there is a tendency with Dutch teachers to consider the number of article errors in their evaluation of essays, Vietnamese teachers seem to overlook article errors completely, which may have various reasons. One possibility may be that Vietnamese teachers tolerate article errors either because they are used to seeing them frequently or they may find them difficult themselves. Another possibility is that because the current emphasis on communication and meaning rather than form and accuracy, Vietnamese teachers may feel that article errors are not problematic enough to pay attention to because they do not often cause communication to break down. But even when article errors are removed completely from four essays and graded by Dutch raters, the grades improved somewhat, but not significantly. This suggests that even teachers who are not used to seeing article errors frequently and who have no difficulty with articles themselves tend to ignore these errors when grading essays holistically. This study also shows that there is no correlation between essay grades and proficiency scores; this is not so surprising considering the fact that proficiency in this study is limited to linguistic competence with a focus on the knowledge of structure and knowledge of lexicon, and writing may be graded more on content rather than form. As far as the relationship between proficiency scores and article errors is concerned, the lack of a correlation confirms the findings in the error analysis. Even the more proficient students still have problems in using the English articles. The Chapter 2 19
results are in conflict with Oller and Redding s remark that the learning of article usage goes hand-in-hand with the development of overall proficiency (1971: 93). What has been found clearly gives the impression that students do not automatically internalize the rules for article use even when they become more proficient. IMPLICATIONS Obviously articles are a pervasive problem for Vietnamese learners of English as they are the most frequent of all errors, which do not decrease as proficiency increases. In other words, they are not learned automatically in the acquisition process. However, it is also clear that holistic essay grades are not much affected by the number of article errors, most probably because article errors are local errors that do not affect meaningful communication much. The question then is should we pay attention to these errors at all and bother teaching the English article system since they are so problematic for Vietnamese learners. It seems to us that at the lower levels, students learning English who will be using English in a local variety do not need much focus on article use because it does not seem to affect meaningful communication. However, the issue is still worth pursuing for advanced learners of English who may have to write professionally for an international public. For, as Berry notes (1991: 252), if a writer keeps making this kind of unimportant error, he or she may create a negative effect on readers. Readers may have the idea that the writer has an imperfect control of the language and his or her integrity may be undermined. Chapter 2 20
Apparently, even university students who will become future teachers of English do not acquire the system through input or study. In the following chapter we will examine the reason why the English article system is so difficult to acquire for students with article-less languages like Vietnamese as their first language. Chapter 2 21